| Metric | naive | lstm | xgboost | cnn |
|---|---|---|---|---|
| diff_of_means | -1.098 | 1.455 | 7.534 | -19.723 |
| ratio_of_sd | 0.830 | 0.809 | 0.857 | 0.902 |
| amplitude_ratio_of_means | 0.539 | 0.581 | 0.529 | 0.687 |
| maximum_error | 0.302 | 0.339 | 0.277 | 0.240 |
| ks_mean_on_coarse_res_with_extremes | 0.587 | 0.333 | 0.436 | 0.316 |
| rainy_hours_ratio_of_means | 0.815 | 0.832 | 0.951 | 0.810 |
| qqplot_mae | 0.029 | 0.032 | 0.018 | 0.039 |
| acf_mae | 0.152 | 0.111 | 0.110 | 0.087 |
| extremogram_mae | 0.109 | 0.058 | 0.083 | 0.065 |
Important: Right now we are only estimating the upper tail
extremogram. Currently we didn’t find a way to estimate the two tales at
the same time. We are using quant = .97
| Metric | lstm | xgboost | naive | cnn |
|---|---|---|---|---|
| diff_of_means | -1.320 | 4.573 | -4.781 | -19.525 |
| ratio_of_sd | 0.849 | 0.878 | 0.894 | 0.913 |
| amplitude_ratio_of_means | 0.591 | 0.526 | 0.552 | 0.666 |
| maximum_error | 0.340 | 0.270 | 0.316 | 0.262 |
| ks_mean_on_coarse_res_with_extremes | 0.227 | 0.387 | 0.378 | 0.196 |
| rainy_hours_ratio_of_means | 0.798 | 0.901 | 0.792 | 0.781 |
| qqplot_mae | 0.033 | 0.019 | 0.028 | 0.046 |
| acf_mae | 0.117 | 0.121 | 0.147 | 0.075 |
| extremogram_mae | 0.069 | 0.090 | 0.117 | 0.057 |
Important: Right now we are only estimating the upper tail
extremogram. Currently we didn’t find a way to estimate the two tales at
the same time. We are using quant = .97
| Metric | lstm | naive | xgboost | cnn |
|---|---|---|---|---|
| diff_of_means | -1.299 | -4.318 | 5.570 | -18.845 |
| ratio_of_sd | 0.870 | 0.901 | 0.890 | 0.914 |
| amplitude_ratio_of_means | 0.590 | 0.551 | 0.533 | 0.663 |
| maximum_error | 0.348 | 0.315 | 0.267 | 0.259 |
| ks_mean_on_coarse_res_with_extremes | 0.292 | 0.474 | 0.407 | 0.277 |
| rainy_hours_ratio_of_means | 0.819 | 0.808 | 0.937 | 0.796 |
| qqplot_mae | 0.031 | 0.026 | 0.017 | 0.044 |
| acf_mae | 0.115 | 0.154 | 0.117 | 0.079 |
| extremogram_mae | 0.078 | 0.124 | 0.094 | 0.063 |
Important: Right now we are only estimating the upper tail
extremogram. Currently we didn’t find a way to estimate the two tales at
the same time. We are using quant = .97
| Metric | xgboost | cnn | lstm | naive |
|---|---|---|---|---|
| diff_of_means | -18.592 | -19.040 | -22.314 | -24.890 |
| ratio_of_sd | 0.933 | 0.934 | 0.892 | 0.884 |
| amplitude_ratio_of_means | 0.618 | 0.745 | 0.703 | 0.599 |
| maximum_error | 0.285 | 0.245 | 0.346 | 0.295 |
| ks_mean_on_coarse_res_with_extremes | 0.340 | 0.183 | 0.206 | 0.442 |
| rainy_hours_ratio_of_means | 0.740 | 0.773 | 0.669 | 0.655 |
| qqplot_mae | 0.037 | 0.034 | 0.046 | 0.053 |
| acf_mae | 0.145 | 0.098 | 0.126 | 0.182 |
| extremogram_mae | 0.082 | 0.046 | 0.049 | 0.103 |
Important: Right now we are only estimating the upper tail
extremogram. Currently we didn’t find a way to estimate the two tales at
the same time. We are using quant = .97
| Metric | cnn | naive | lstm | xgboost |
|---|---|---|---|---|
| diff_of_means | -17.250 | 19.995 | 20.644 | 30.082 |
| ratio_of_sd | 0.887 | 0.709 | 0.711 | 0.707 |
| amplitude_ratio_of_means | 0.651 | 0.451 | 0.491 | 0.416 |
| maximum_error | 0.259 | 0.281 | 0.333 | 0.248 |
| ks_mean_on_coarse_res_with_extremes | 0.248 | 0.569 | 0.263 | 0.461 |
| rainy_hours_ratio_of_means | 0.880 | 0.995 | 1.026 | 1.204 |
| qqplot_mae | 0.043 | 0.037 | 0.044 | 0.041 |
| acf_mae | 0.097 | 0.171 | 0.115 | 0.128 |
| extremogram_mae | 0.084 | 0.161 | 0.081 | 0.127 |
Important: Right now we are only estimating the upper tail
extremogram. Currently we didn’t find a way to estimate the two tales at
the same time. We are using quant = .97
| Metric | naive | lstm | xgboost | cnn |
|---|---|---|---|---|
| diff_of_means | 6.056 | 7.760 | 17.315 | -20.448 |
| ratio_of_sd | 0.787 | 0.790 | 0.784 | 0.920 |
| amplitude_ratio_of_means | 0.511 | 0.562 | 0.487 | 0.694 |
| maximum_error | 0.275 | 0.319 | 0.240 | 0.247 |
| ks_mean_on_coarse_res_with_extremes | 0.565 | 0.269 | 0.461 | 0.304 |
| rainy_hours_ratio_of_means | 0.874 | 0.916 | 1.060 | 0.851 |
| qqplot_mae | 0.029 | 0.033 | 0.025 | 0.036 |
| acf_mae | 0.154 | 0.103 | 0.108 | 0.092 |
| extremogram_mae | 0.106 | 0.048 | 0.078 | 0.061 |
Important: Right now we are only estimating the upper tail
extremogram. Currently we didn’t find a way to estimate the two tales at
the same time. We are using quant = .97
| Metric | cnn | naive | lstm | xgboost |
|---|---|---|---|---|
| diff_of_means | -16.461 | 18.193 | 18.707 | 28.620 |
| ratio_of_sd | 0.898 | 0.734 | 0.733 | 0.728 |
| amplitude_ratio_of_means | 0.644 | 0.465 | 0.500 | 0.427 |
| maximum_error | 0.266 | 0.276 | 0.325 | 0.252 |
| ks_mean_on_coarse_res_with_extremes | 0.261 | 0.473 | 0.255 | 0.409 |
| rainy_hours_ratio_of_means | 0.870 | 0.979 | 1.012 | 1.192 |
| qqplot_mae | 0.044 | 0.036 | 0.043 | 0.040 |
| acf_mae | 0.088 | 0.153 | 0.111 | 0.109 |
| extremogram_mae | 0.069 | 0.135 | 0.078 | 0.106 |
Important: Right now we are only estimating the upper tail
extremogram. Currently we didn’t find a way to estimate the two tales at
the same time. We are using quant = .97